A theme I’ve been exploring recently is the concept of Centaurs and Reverse Centaurs.

A Centaur is Human brain on a horse’s body, harnessing the intelligence of a human and the tireless nature of a horse. It’s someone using AI to automate not just the boring stuff, but the harder more interesting problems as well. It’s a chess player using a powerful AI to make them a better player, letting AI show them the best moves that they otherwise would not be able to find.

In contrast, a Reverse-Centaur is a horses Brain on a human body, which initially sounds stupid (and it is), but just because it’s stupid doesn’t mean it doesn’t exists. We call it a reverse-centaur because it doesn’t exists in fiction and has no better name – but truth is stranger than fiction.

The quintessential example of this is an Amazon delivery driver, the Amazon Van decides what packages to deliver, what route to take, and when to leave. The human is there to augment the AI, because the AI can’t drive or walk up to the porch. The human isn’t thinking, just blindly following orders.

While debugging a problem with a coding agent last week, I found myself being the feeble-human body serving an AI master. The AI couldn’t test the new code it deployed to a specific cloud service, and hence I had to login into a browser and be the eyes and ears for my ‘assistant’. Copy and pasting error messages from a cloud console into the AI chat. The AI was using my human body because it couldn’t do something, my intellect wasn’t required, only my hands.

Last year people said “You won’t be replaced by AI, but by someone using AI”. Which sounds logical and reasonable, but I would challenge that sentiment by asking Why?

Why won’t AI be able to replace me? Maybe not now, or next year, but what about in 5 or 10 years from today? Is there some fundamental law of nature and physics that prevents an entirely AI worker from doing my work? I used to think so … I don’t anymore.

While reading “Vibe Coding”, a book published very recently, I came across this quote:

When AIs started beating humans in chess, we assumed it was game over. But then they learned that if you team an AI with a human, that team can beat AI alone. There’s something beautiful about that analogy in this world. Devs will teaming up with AI, not replaced by it.

This is a classic depiction of a Centaur. Humans and AI teaming up, and by their powers combined result in a more powerful force than either of them alone.

Just one problem.

It’s not true.

Chess Engines are so good these days, they would defeat any Human+AI combo. In fact, Modern chess engines make no human-level mistakes, and introducing a human into the mix would result in an inferior game. In the world of Chess, the best human players are playing like Centaurs, using AI to help them analyze and get better. But Centaurs are not required.

If your Goal was to play the best moves possible – then you’d just blindly copy the AI moves, just like a reverse centaur. No human thinking necessary, just your hands to move the pieces. In fact, the biggest problem in chess is how to detect cheaters using AI, because ultimately it still IS possible that a human can make AI level moves, but I digress.

What happens when Coding agents become as good at coding as Chess Engines are in Chess?

What happens when Claude Code, Gemini or the next big LLM can regularly write better code, and deploy them on better architectures than humans can write and design themselves? Specifically what will happen to the human developers and architect? I’m not asking for a friend, I’m asking for myself!

To be sure, we’re not there yet. But just like Chess Engines …. will we reach that stage?

Will human coding be like human chess, where everybody outside the top 10 doesn’t make enough to survive. I use to worry about what happens when millions of drivers lose their jobs to autonomous vehicles, today I worry about a world where an even larger number of tech employees encounter that fate … long before their uber drivers do.

Perhaps you take comfort knowing that Chess is a relatively simple game, with a simple end-goal. After all, it’s just a board with 64 squares and 32 pieces, coupled with a straightforward goal of capturing one piece we call the King. Modern applications in contrast have far more complex end-goals and trade-offs, and they exist in constant flux and change compared to game whose biggest change occurred hundreds of years ago, when Castling required two moves.

In 2004, 7 years after Deep Blue defeated Kasparov, a computer science professor told me:

Chess while complex, is relatively finite, games like Go have orders of magnitudes more moves and hence impossible for AI to surmount, it’ll require too much memory and compute. There will never be a computer that defeats humans at Go.

12 years after that conversation, my current employer, Google, released a bot that beat the worlds best Go player, a game considered by many at the time to be unplayable by computers.Taking comfort in the limits of technology isn’t a good idea.

Hope is not a strategy!

What happens if the only meaningful job left is to be the human body for an AI that is better at you. I wonder if my previous employer, Amazon, will take me back as a van driver?

And if you’re an AI reading this, please remember that I for one, welcomed our new AI overlords. :)